How to check Amazon product demand properly

Use first-party data before modelled data

There are two kinds of demand signal and the difference is important. First-party data comes from Amazon itself. Modelled data is inferred, usually from Best Sellers Rank, by a third-party tool. Both are useful; only one is a fact.

Product Opportunity Explorer (free, in Seller Central)

Amazon's own view of customer demand by niche: search volume, growth trend, how concentrated the top products are, average price and return rates. This is the single most underused resource in Amazon research, largely because nobody earns a commission recommending it.

Brand Analytics search-term reports (free, with Brand Registry)

What customers actually search, which ASINs they click, and which they buy. Only available once you have a registered brand, so not a pre-launch tool — but worth knowing it exists before you assume third-party estimates are the ceiling.

The search bar itself (free, no account)

Amazon's autocomplete is ranked by real search behaviour. Type your product and read the suggestions — they tell you what people actually search for, in their words, and frequently surface a more specific variant with its own demand that you had not considered.

How to read modelled sales estimates

Most research tools show estimated monthly sales, and nearly all of them derive the figure from Best Sellers Rank. Use those numbers to tell a market of hundreds of units from one of thousands. Two products estimated at 1,847 and 1,610 a month are, for decision purposes, the same. Our Amazon sales estimator guide covers how the curves are built, where they break, and the free first-party figures to check an estimate against.

SellRadar deliberately does not publish sales estimates. It is a real gap if estimates are how you filter, and several tools do that job well.

Free demand signals that do not require a tool

1. Review velocity on the top listings. Compare review counts against how long listings have been live. Fast accumulation means steady sales; a five-year-old listing with 200 reviews means a slow market.
2. Depth of the page. If twenty listings all sustain meaningful review counts, the market supports twenty sellers. If only the top three have any traction, demand concentrates at the top and position four is a much worse place to be than it looks.
3. Autocomplete breadth. Many distinct suggested variants means many distinct searches, which means a market with room in it.
4. Recency of reviews. Open the top listing and sort reviews by most recent. A steady flow of recent reviews is live demand. A gap of months is a warning regardless of the total count.
5. Google Trends on the product term. Not Amazon-specific, but it will catch a category in decline and it will show you seasonality shape at a glance.
6. Question volume. Customer questions on established listings indicate engaged buyers, and the questions themselves frequently name the unmet need.

None of these are precise. Together they answer the question that actually matters at this stage — is there a real market here, roughly how big, and is it growing or dying — which is all demand data needs to tell you.

Check whether you are looking at a spike

This is where snapshot demand data goes badly wrong, and it is the most expensive demand mistake because it inverts the answer rather than blurring it.

A product researched in its peak season looks extraordinary. The same product in February looks dead. Sellers commit inventory based on a peak reading, receive stock after the season closes, and spend the year storing it.

• Check price and rank history — Keepa's basic charts are free and answer this in under a minute.
• A rank line with an annual spike tells you everything a snapshot cannot.
• If demand is seasonal, that is not automatically disqualifying. It means your cash cycle and inventory planning have to match it, and your first year has one shot rather than four quarters.
• For a first product, prefer flat demand. You want the chance to be wrong and recover.

Why demand is the easy half

Here is the thing worth internalising. Products almost never fail because nobody wanted them. They fail because the seller could not take a share of demand that was already there.

If a product has twenty listings on page one with real review counts, demand is proven — twenty sellers are making sales. What is unproven is whether there is room for a twenty-first, and that depends on how those twenty positions are defended: how review depth is distributed, how many holders are generic versus branded, whether listing quality is beatable, whether the price band survives the fee stack.

So confirm demand quickly and cheaply, then spend your real effort on the competitive read. Sellers who invert this end up with precise demand estimates for markets they cannot enter.

SellRadar handles the second half: type a product and it reads the top 20 organic listings, flags each competitor's weaknesses, and returns LAUNCH, WATCH or SKIP with the reasoning, written for your selling model, across thirteen marketplaces. Three a month free, no card.

Not all demand is worth the same

Demand is usually discussed as a quantity. It is also a quality, and two markets with identical volume can be worth very different amounts to a new seller.

The first row is the one most often missed and the most consequential. If a market's volume concentrates almost entirely in the top three listings, then entering at position eight earns you a fraction of what the headline demand implies — and the headline is what most tools report.

You can check it without any tool: look at whether review counts and recency thin out sharply after the first few listings, or continue meaningfully down the page. A market that sustains twenty sellers is a market with room for a twenty-first. One that sustains three is not, however large the total looks.

Frequently asked questions